Low-Carbon Dispatch Strategy for Virtual Power Plant Based on Bagging Ensemble Learning Forecasting Method

Virtual power plants (VPPs) struggle with peak-shaving due to unreliable distributed generation (DG) data and the intermittent nature of wind and solar power, impacting carbon emissions. This paper proposes a low-carbon dispatch strategy for VPPs based on a Bagging ensemble forecasting model. Firstly, a VPP dispatch framework based on DG output and load forecasting is established, and a Bagging ensemble forecasting model with feature selection is proposed. Secondly, the control characteristics of flexible industrial, commercial, and residential loads are analyzed to establish a control cost model, which is then integrated with the balanced relationship between wind, photovoltaic, and energy storage, as well as the carbon emission mechanism, to formulate a low-carbon dispatch model. Finally, the simulation results show that the proposed strategy can improve the accuracy of DG output and load forecasting in complex feature environments while realizing the interaction between source-grid-load and reducing the generation and operation costs of the grid.